Genetic Diversity and Drivers of Genetic Differentiation of Reaumuria Soongorica of the Inner Mongolia Plateau in China

Plant Ecol (2015) 216:925–937 DOI 10.1007/s11258-015-0479-3

Genetic diversity and drivers of genetic differentiation of Reaumuria soongorica of the Inner Mongolia plateau in China

Jiuyan Yang . Samuel A. Cushman . Xuemei Song . Jie Yang . Pujin Zhang

Received: 7 October 2014 / Accepted: 30 April 2015 / Published online: 15 May 2015 Ó Springer Science+Business Media Dordrecht (outside the USA) 2015

Abstract We quantified genetic diversity and gene change on this keystone foundation species and flow among eight populations of Reaumuria soon- devising effective strategies to utilize it in restoration gorica in Inner Mongolia, China. Our results showed efforts to ameliorate ongoing desertification in the that genetic differentiation of R. soongorica across the region. Genetic diversity was highest in the western Inner Mongolian plateau is primarily clinal in nature part of the sampled population, perhaps indicating that and is driven primarily by differential landscape this region has historically harbored the highest resistance across areas with changing patterns of effective population size of the species or may have seasonal precipitation, perhaps as a result of differen- served as the source of recent range expansion to other tial timing of reproductive phenology along precipita- parts of the sampled range which exhibited lower tion gradients. Finding that seasonal patterns of genetic diversity. Understanding the ecological dri- precipitation, and not temperature, drive population vers of these relationships might be critical to connectivity and gene flow may have important resolving the causes of the geographical pattern of implications for predicting the effects of climate diversity, and could be important in understanding the ecology of the species sufficiently to anticipate climate change effects and effectively implement manage- Communicated by Siegy Krauss. ment strategies to restore the species and combat desertification. Electronic supplementary material The online version of this article (doi:10.1007/s11258-015-0479-3) contains supplementary material, which is available to authorized users. Keywords Reaumuria soongorica Á Reciprocal causal modeling Á Landscape genetics Á Climate Á J. Yang Á X. Song Á J. Yang STRUCTURE analysis School of Life Sciences, Inner Mongolia University, 235 West University Blvd, Hohhot 010021, China

S. A. Cushman (&) Introduction USDA Forest Service/Rocky Mountain Research Station, 800 E Beckwith, Missoula, MT 59801, USA e-mail: [email protected] Spatially heterogeneous natural selection and genetic drift can lead to population divergence even in the P. Zhang presence of gene flow (Coyne and Orr 2004; McKin- Inner Mongolia Prataculture Research Center, Chinese non et al. 2004). Reaumuria soongorica in the Academy of Science/Inner Mongolia Academy of Agricultural and Animal Husbandry Sciences, Mongolian Plateau, China, provides an excellent Hohhot 010031, China opportunity to investigate patterns of genetic 123 926 Plant Ecol (2015) 216:925–937 differentiation of peripatric populations along steep is recognized as the most important grazing land in the climatic gradients given that the species spans multi- Inner Mongolian region (Inner Mongolia and Ningxia ple climatic zones and vegetation types, across a broad Comprehensive expedition of Chinese academy of range of moisture, temperature and soils, and exhibits sciences 1985;Ma1989;Wu1980). extensive adaptability and a broad ecological niche. R. Analysis of gene flow and genetic differentiation of soongorica first occurs in the fossil record in the R. soongorica along this climatic gradient may allow Cretaceous and Paleocene (Hou 1983). In the early identification of the ecological and climatic drivers of stage of the Tertiary, there was a long arid period, recent microevolution of this species. Additionally, during which the R. soongorica community developed understanding the factors that control gene flow and (Yong 1990; Zhang 2005). In the desert areas of genetic diversification of R. soongorica may help central Asia, the R. soongorica formation is one of the clarify the evolutionary and ecological processes most important community types, and is the zonal behind its keystone ecological role in plant commu- desert community that is most widespread and covers nities of the Mongolian plateau. Due to its environ- the largest area in the Mongolian plateau. R. soon- mental benefit and utilization value, many studies of R. gorica is widely distributed across a broad range of soongorica have been conducted on its biological vegetation zones, spanning from typical desert, steppe characteristics (Liu et al.1982), morphology (Cui desert, desert steppe, and typical steppe from west to 1988), anatomy (Xiao et al. 2006), physiology (Li east in the Mongolian plateau. In a typical desert et al. 2005a, b), genetic diversity (Zhang et al. 2008), community, R. soongorica is found in all soil types seed germination and breeding conditions (Zeng et al. and is a keystone foundation species for community 2004), and grazing and cultivating (Wang et al. 2002a, formation. In steppe desert, it is an adventitious b) all of which have contributed to a substantial community, which primarily occurs on gravelly, sandy knowledge of the basic biology and management of R. and low saline soils. In the western part of Mongolian soongorica. However, knowledge of the molecular plateau, R. soongorica often is the principal dominant ecology of this species is still extremely limited. species, while eastward its dominance declines as There is a lack of knowledge about genetic aridity decreases, where it is predominantly found diversity and differentiation within and among along saline lowlands and the peripheries of salty lakes populations of R. soongorica at broad scales. Recent- (Editorial Committee of Flora Reipublicae Popular ly, Li et al. (2012) identified three major chloroplast Sinicae, Chinese Academy of Science 1990; Inner haplotype clades across the species’ range that may be Mongolia and Ningxia Comprehensive expedition of associated with the effects of the Qinghai-Tibet Chinese academy of sciences 1985;Li1990). Given Plateau and development of desert ecosystems during this broad ecological amplitude, the R. soongorica the last glacial age in western China. In addition, Qian formation is the most widespread, and the most et al. (2008) found reasonably high genetic diversity ecologically and economically important zonal com- and strong evidence of isolation by distance across a munity type on the Inner Mongolia plateau. small portion of the species range. However, to date no Reaumuria soongorica is a xerophytic and salt- studies have compared multiple hypotheses about the tolerant undershrub, with strong resistance to distur- effects of climatic gradients on genetic differentiation bances such as soil erosion (Inner Mongolia and and gene flow of this species. Given the very high Ningxia Comprehensive expedition of Chinese acade- importance of the species in efforts to ameliorate my of sciences 1985; Liu et al. 1982;Ma1989;Wu desertification and its importance to livestock produc- 1980). Indeed, it has an extremely strong ability to tion, it is essential to understand the climatic controls stablilize shifting sand, making it extremely important on the species’ distribution, gene flow and genetic in reducing effects of desertification and over grazing, differentiation. In particular, proactive management which have been major environmental problems in the depends on understanding the potential effects of region over the past 50 years (Li et al. 2008; Liu et al. climate change on the occurrence, productivity and 1982; Ma and Kong 1998). In addition, R. soongorica population connectivity of this species. The distribu- is an important forage species, and is additionally tion characteristics of R. soongorica in the Inner valuable given that it is a major source of salt for Mongolia plateau provide an opportunity to investi- livestock, and grassland dominated by R. soongorica gate the relationships between genetic diversity, gene 123 Plant Ecol (2015) 216:925–937 927

flow and environmental gradients. In addition, because mountains to the north. The plateau includes the Gobi of recent climatic changes and long-term overgrazing, Desert as well as extensive dry steppe regions. It has an R. soongorica has declined in extent and dominance, elevation range of approximately 700–1500 m, with leading to ecological decline of the plant communities the lowest point in Hulunbuir and the highest point in that depend on it, accelerating desertification and the Altai Mountain range. economic loss. Therefore, research to understand the The current study was conducted on the portion of ecology and enhance the restoration of this species is the Mongolian Plateau within the Chinese province of of high importance. Inner Mongolia. Most of Inner Mongolia is a plateau Investigations of genetic diversity, genetic differ- averaging around 1200 m in altitude (700–1400 m entiation and gene flow can provide a molecular basis range), with a topographical cline rising from the for understanding the biological characteristics, evo- southwest to the northeast. The soils are characterized lutionary history and adaptive potential of this species, by extensive loess and sand deposits. Soil types as well as its sensitivity to climate change and generally change from chestnut soil, to brown soil, to potential strategies to enhance its restoration. Neutral gray desert soil, and gray brown desert soil from east nuclear DNA markers are considered to be the most to west. Inner Mongolia has a wide variety of regional suitable means for estimating genetic diversity be- climates, with aridity increasing markedly from east to cause of their abundant polymorphism and the fact that west. The climate is characterized by a four-season, they are independent of environmental conditions monsoon climate, with long, cold, and dry winters, (Gupta et al. 1994; Zietkiewicz et al. 1994). ISSR short, dry spring, and warm, relatively humid sum- (Inter-simple sequence repeat) can provide more mers, except in the western parts of the region where abundant polymorphisms and more reliable and the monsoon pattern does not reach. Mean annual reproducible bands relative to other DNA markers temperature and temperature extremes increase from (Qian et al. 2001; Weising et al. 1995), and therefore east to west (C10 °C active accumulated temperature have been widely used in detecting genetic diversity is 1300–3700 °C). Annual precipitation is and geographic population variation, especially in 550–10 mm, and the moisture index, which is a distinguishing genetic variation below species levels function of both temperature and precipitation, is (Song et al. 2008). 1.0–0.01, both decreasing from east to west. The In this study, we used ISSR markers to analyze the natural vegetation changes in a gradient from mesic patterns of genetic diversity within and among popula- grassland and steppe in the east to desert in the west, tions of R. soongorica, investigate the characteristics along a chronosequence of meadow steppe (sub- that enable the wide distribution of the species across the humid), typical steppe (semi-arid), desert steppe Inner Mongolia plateau, and quantify the relationships (drought), steppe desert (drought), and typical desert between genetic diversity, genetic differentiation, and (extreme drought). Grazing is the dominant economic climatic and environmental gradients. Our goal is to activity (Inner Mongolia and Ningxia Comprehensive elucidate relationships between ecological gradients expedition of Chinese academy of sciences 1985;Wu and genetic structure to provide guidance for biodiver- 1980). sity protection in the Inner Mongolia plateau. Plant material

Materials and methods We chose eight sites along the climatic gradient from east to west across the Inner Mongolia plateau at Study area which we collected tissues from R. soongorica. These sites were located in four different vegetation zones The Mongolian Plateau is the part of the Central Asian including typical steppe, desert steppe, steppe desert, Plateau lying between 87°400–122°150N and 37°460– and typical desert, and included several sampling 53°080E, with an area of approximately 2600,000 km2. locations in transitional zones (Fig. 1; Table 1). It is bounded by the Greater Hinggan Mountains in the Across the eight sampling sites, both the average east, the Yin Mountains to the south, the Altai annual precipitation and moisture index decrease Mountains to the west, and the Sayan and Khentii gradually. 123 928 Plant Ecol (2015) 216:925–937

Fig. 1 Collection sites of Reaumuria soongorica on the Inner Mongolia plateau. Sampling sites are given in Table 1

In each sampling site, 30 individuals of R. soon- Research PTC-100 thermocycler (Bio-rad, Waltham, gorica were randomly selected for sampling, produc- MA, USA) programmed for an initial denaturation ing a total sample of 240 individuals. Fresh leaves temperature of 94 °C for 5 min and then 40 cycles of collected from each selected individual were stored 45 s at 94 °C, 45 s at 52 °C, 1.5 min at 72 °C, with a with silica gel in sealed plastic bags for later DNA 5-min final extension at 72 °C. The amplification extraction. The details on R. soongorica populations products were separated on 2.5 % (w/v) agarose gels and the sampling sites are shown in Fig. 1 and Table 1. with 0.5 g/L ethidium bromide electrophoresed in 0.59 TBE buffer (0.45 mM Tris–borate, 0.01 mM DNA extraction EDTA, pH 8.0) at 5 V/cm for 2 h. DNA fragments were visualized and photographed under ultraviolet Total genomic DNA was extracted following the light with WFH-701Type UV analyzer. Molecular CTAB (cetyltrimethylammonium bromide) protocol weights were estimated from a 100 bp DNA ladder. (Doyle 1999). The overall quantity and quality of extracted DNA were determined in 0.8 % (w/v) WorldClim climate layers agarose gels. The total DNA extracted for each sample was diluted to 20 ng/ll and stored at -20 °C. Bioclimatic variables derived from the monthly temperature and rainfall values are often used in eco- PCR amplification logical niche modeling (Iverson and Prasad 2002; Rehfeldt et al. 2006) or analyses of influences of Nuclear DNA was polymerase chain reaction (PCR)- climatic gradients on gene flow (Wasserman et al. amplified using ISSR primers from the University of 2010). We selected 18 bioclimatic variables for British Columbia (Canada). Following an initial analysis from the WorldClim database (Hijmans screening of 100 random primers, 14 that gave clear et al. 2005; Table 2). WorldClim climatic layers are reproducible fragments were selected for further derived from monthly temperature and rainfall clima- analysis (Table 2). PCR was carried out in a total tologies and represent biologically meaningful vari- volume of 20 ll consisting of 10 llof29 Taq PCR ables for characterizing species ranges (Nix 1986). Master Mix, 1 ll of primer, 1 ll of template DNA, and The WorldClim database was developed using a double-distilled water. All the reagents were pur- global network of weather stations whose data were chased from Sangon Biotech (Shanghai, China) Co., interpolated to monthly climate surfaces at 1 km Ltd. The amplifications were performed in a MJ spatial resolution using a thin-plate smoothing spline

123 Plant Ecol (2015) 216:925–937 929

Table 1 Ecological characteristics of the eight sampling sites where genetic data on Reaumuria soongorica were recorded on the Inner Mongolia plateau for this study Sampling Population Vegetation type Moisture Type of soil WorldClim mean annual WorldClim mean annual sites identiﬁer index temperature (°C) precipitation (mm)

1 RSI Typical steppe 0.3–0.6 Dark chestnut 2.1 266 soil 2 RSII Typical steppe 0.3–0.6 Dark chestnut 2.4 224 soil 3 RSIII Typical steppe— 0.13–0.3 Light 3.5 195 desert steppe chestnut soil 4 RSIV Desert steppe 0.13–0.3 Brown soil 4.7 173 5 RSV Desert steppe— 0.13–0.3 Brown soil 4.5 197 steppe desert 6 RSVI Steppe desert 0.05–0.13 Brown soil 4.4 153 7 RSVII Steppe desert— 0.05–0.13 Gray desert 7.3 102 typical desert soil 8 RSVIII Typical desert \0.02 Gray brown 8.9 33 desert soil Vegetation type, moisture index, and soil type as deﬁned in Inner Mongolia and Ningxia Comprehensive expedition of Chinese academy of sciences (1985) Vegetation type, moisture index, and type of soil were cited from Vegetation of Inner Mongolia

Table 2 List of bioclimatic Acronym Description variables used in analysis of gene ﬂow G1 Annual mean temperature G2 Mean diurnal range (mean of monthly (max temp–min temp) G4 Temperature seasonality (SD 9 100) G5 Max temperature of warmest month G6 Min temperature of coldest month G7 Temperature annual range (BIO5–BIO6) G8 Mean temperature of wettest quarter G9 Mean temperature of driest quarter G10 Mean temperature of warmest quarter G11 Mean temperature of coldest quarter G12 Annual precipitation G13 Precipitation of wettest month G14 Precipitation of driest month G15 Precipitation seasonality (coefﬁcient of variation) G16 Precipitation of wettest quarter G17 Precipitation of driest quarter G18 Precipitation of warmest quarter G19 Precipitation of coldest quarter algorithm with latitude, longitude, and elevation as seasonality (e.g., annual range in temperature and independent variables (Hijmans et al. 2005). The precipitation) and extreme or limiting environmental bioclimatic variables represent annual trends (e.g., factors (e.g., temperature of the coldest and warmest mean annual temperature, annual precipitation), month, and precipitation of the wet and dry quarters). 123 930 Plant Ecol (2015) 216:925–937

Data analysis value FST \ 0.05 represents almost no genetic differentiation among populations. When 0.05 \ FST Measures of genetic diversity \ 0.15 there is a moderate amount of differentiation among populations. FST values between 0.15 and 0.25 Only reproducible and unambiguous ISSR bands were represent highly differentiated among populations, and scored for presence (1) or absence (0). Data were FST [ 0.25 indicates very high differentiation among compiled into a binary data matrix using MS Excel. populations (Excoffer et al. 1992;Wright1931). Population genetic parameters were analyzed using POPGENE version 1.31 (Yeh et al. 1999)todetermine Evaluating drivers of genetic structure the percentage of polymorphic bands (PPB), Nei’s (1973) gene diversity (H), Shannon’s information index The predominant analytical approach to associate (I*), total genetic diversity (Ht), genetic diversity within landscape patterns with gene flow processes is based populations (Hs), genetic diversity among populations on pair-wise calculation of cost distances, which are (Dst), coefficient of genetic differentiation (Gst), Nei’s then correlated with pair-wise genetic distances among (1972) genetic distance (D) between pairs of popula- the same individuals with methods such as Mantel and Ã tions, and genetic identity (I). Gene flow Nm was partial Mantel tests (Mantel 1967; Smouse et al. 1986). Ã calculated according to the formulae: Nm ¼ 0:5 There has been controversy in the literature about the ðÞ1 À Gst =Gs. appropriateness of Mantel testing in landscape genetics. Recently, Legendre and Fortin (2010) clarified this confusion, and note, that while distance-based regres- STRUCTURE analysis of genetic structure sion approaches, such as the Mantel test, have lower power than traditional linear models and tend to The population genetic structure of R. soongorica was underestimate the true magnitude of a relationship, quantified using the Bayesian admixture procedure partial Mantel testing remains the appropriate frame- implemented in the program STRUCTURE (Falush work when the hypotheses are explicitly defined in et al. 2003; Pritchard et al. 2000). The program infers the terms of distance matrices, as they are in landscape genetic structure by assuming a model in which there are genetic analyses testing effects of landscape resistance K (known or unknown) genetic clusters, and individual on neutral genetic differentiation. genotypes are probabilistically assigned to genetic Cushman et al. (2006) proposed a causal modeling clusters, or jointly to two or more genetic clusters if framework to assist in model selection and increase their genotypes indicate that they are admixed (Pritchard the likelihood of identifying the true driver of genetic et al. 2000). Analyses were carried out with 100,000 isolation. This approach involves identifying the most iterations, following a burn-in period of 100,000 supported resistance hypothesis among a range of iterations, in order to provide insights into the genetic alternative resistance models (based on statistical structure by assigning individuals to K (which ranged significance) and then using partial Mantel tests to from 1 to 8) using only genotypic data. The most determine whether it meets the statistical expectations appropriate number of genetic clusters was determined of a causal model relative to alternative models of based on ad hoc statistic, DK, which evaluates the isolation by distance or isolation by barrier. Cushman second order rate of change in the likelihood between and Landguth (2010) evaluated the power of this successive K values (Evanno et al. 2005). Three framework and found that the method performs well in independent runs were carried out for each K value in identifying the drivers of genetic differentiation in a order to assess the consistency of results. case study complex landscape, and rejecting incorrect and correlated alternatives. Analysis of molecular variance (AMOVA) Recently, Cushman et al. (2013) found that partial Mantel tests have very low Type II error rates, but Analysis of molecular variance for the eight R. elevated Type I error rates. This leads to frequent sonngorica populations was conducted using software identification of support for spurious correlations Arlequin 3.11. The fixation index (FST) is widely used between alternative resistance hypotheses and genetic to estimate the genetic variation among populations. An distance, independent of the true resistance model. As 123 Plant Ecol (2015) 216:925–937 931 a result Cushman et al. (2013) suggested several Results changes to the original causal modeling framework developed by Cushman et al. (2006), based on the Genetic polymorphism relative support of the causal modeling diagnostic tests, rather than formal hypothesis testing. This A total of 316 bands were amplified using 14 ISSR improved method reduces the problem of false primers with the band size ranging from 100 to positives (Type I errors) observed by Meirmans 2000 bp (Table 2). On average, 22.57 bands were (2012) and Amos et al. (2012). used per primer. All 316 loci were polymorphic (the We employ this relative support method in our percentage of polymorphic bands, PPB = 100 %), analysis. Specifically, our analysis is based on com- indicating a high level of polymorphism in genome of puting two partial Mantel tests for each combination of R. soongorica (Table 2). alternative resistance hypotheses. For each alternative model, these two tests include: computing the partial Genetic diversity Mantel correlation between genetic distance and the focal hypothesis, partialling out each of the other The PPB in each R. soongorica population varied from hypotheses (Test 1), and the partial Mantel correlation 60.13 % to 75.63 %, with RSVII [ RSV = RS- between genetic distance and each of the alternative VIII [ RSIII [ RSI [ RSVI [ RSIV [ RSII. Nei’s hypotheses, partialling out the focal hypothesis (Test gene diversity (H) ranged from 0.1772 to 0.2324, 2). The support for a particular hypothesis relative to a with RSVII [ RSVIII [ RSI [ RSV [ RSIII [ RS- particular alternative is measured by the magnitude of VI [ RSII [ RSIV. Shannon’s information index the difference between these two tests. A large (I) varied from 0.2720 to 0.3554, with RSVII [ RS- positive value for this difference indicates strong VIII [ RSI [ RSV [ RSIII [ RSVI [ RSII [ R- support for the first hypothesis independent of the SIV (Table 3). These three variables all show the second, and little or no support for the second same pattern, with genetic diversity highest for RSVII independent of the first. A well supported hypothesis (desert region), RSVIII (desert region), and RSI will have large positive values for this difference with (steppe region), intermediate for RSV (desert all alternative models. In our analysis, we computed steppe—steppe desert), RSVI (steppe desert), and the full matrix of the difference in support for all RSIII (typical steppe—desert steppe). The level of combinations of 19 alternative resistance hypotheses genetic diversity was lowest for RSII (typical steppe) (18 WorldClim climate hypotheses and Isolation by and RSIV (desert steppe). These results suggest that Distance) plus the two supported STRUCTURE genetic diversity does not increase or decrease mono- cluster assignments (K = 3 and K = 7; e.g., Cushman tonically with aridity. For all measures of genetic et al. 2014). We evaluated which of the selected diversity assessed, we found the highest genetic hypotheses had the most support relative to the other diversity of the species was in the western desert well supported hypotheses. habitats.

Table 3 Analysis of genetic diversity across 316 loci for eight populations using ISSR markers, with 22.57 loci per primer Population identiﬁer Percentage of polymorphic loci % (PPB) Nei’s gene diversity (H) Shannon’s information index (I)

RSI 64.87 0.2096 0.3163 RSII 60.13 0.1805 0.2762 RSIII 66.46 0.2047 0.3116 RSIV 60.44 0.1772 0.2720 RSV 68.04 0.2071 0.3161 RSVI 62.03 0.1906 0.2920 RSVII 75.63 0.2324 0.3554 RSVIII 68.04 0.2110 0.3216

123 932 Plant Ecol (2015) 216:925–937

Genetic differentiation and gene ﬂow

The Nei’s coefficient of genetic differentiation (Gst) for populations of R. soongorica was 0.1931, indicating that 19.31 % of total variation occurred among the populations and 80.69 % occurred within populations, Ã and gene flow ðNmÞ was 2.0899, indicating substantial genetic migration among populations. AMOVA showed that 26.23 % of total variation occurred among the populations and 73.77 % occurred within populations. Fixation index for populations of R. soongorica was 0.26233 ([0.25) suggesting there is high genetic differentiation among populations. However, the value of gene flow ðNÃ Þ for the populations was Fig. 3 Results of Bayesian analysis for 8 populations of R. m sonngorica from STRUCTURE under admixture model for Nm = 2.0899, indicating a moderate level of genetic a K = 7 and b K = 3. 1 RSI, 2 RSII, 3 RSIII, 4 RSIV, 5 RSV, 6 exchange between populations, which may prevent RSVI, 7 RSVII, 8 RSVIII genetic differentiation among populations due to genetic drift. These results suggest partial isolation of populations and varying levels of gene flow, which are both located in the ecological transition suggesting clinal population structure. zone. In the K = 3 solution, RSVIII individuals were Bayesian analyses assigned into a single cluster. The clustering indicated similar genetic structure of individuals from RSV, Cluster results of STRUCTURE revealed three or RSVI, and RSVII. The high degree of admixture seven genetic clusters confirmed by calculation of indicated that there is considerable genetic exchange DK (Fig. 2). In the K = 7 solution, all populations among the eight populations, with particularly strong were assigned to single clusters by themselves, except gene flow among RSI, RSII, RSIII, RSIV, and RSV for population RSV and RSVII which formed one (Fig. 3b). genetic cluster (Fig. 3a). This suggests that the genetic structure of RSV and RSVII are similar, indicating Reciprocal causal modeling with Mantel tests relatively high gene flow among these populations Reciprocal causal modeling across 19 resistance hypotheses and two structure groupings revealed that a single hypothesis had strong support based on the matrix of difference in support between partial Mantel test 1 and partial Mantel test 2 (Fig. 4). Specifically, interseasonal variability in precipitation (G15) was well supported compared to all alternative hypotheses (positive values in columns associated with these hypotheses in Fig. 4). In contrast, none of the other 17 alternative resistance hypotheses or isolation by distance were supported, and none had any support independent of G15 (negative or zero values in the rows associated with these hypotheses in Fig. 4). There was some residual support for the two STRUCTURE cluster solutions (K = 3 and K = 7) Fig. 2 Mean LnPr (K) over three runs for each K value and DK obtained with Bayesian analyses implemented in the independent of interseasonal variability in precipita- software STRUCTURE tion (G15), indicating some genetic structure not 123 Plant Ecol (2015) 216:925–937 933

kilometers, with sampling ‘‘sites’’ representing clustered sampling along this cline. Given the Schwartz and McKelvey (2009) findings, we would expect STRUCTURE to identify clusters from this sampling regime even if the underlying genetic structure was continuous and clinal in nature. Indeed, this was one of the main motivations for our use of comprehensive analysis with reciprocal causal modeling to directly compete clustering, isolation by distance and a range of isolation by resistance hypotheses. Our analysis was able to evaluate the relative and independent support among all candidate hypotheses and confirmed both a dominant pattern of clinal variation in genetic differentiation associated with gradients of seasonal precipitation and also some weaker residual structure Fig. 4 Matrix of difference in support for column hypotheses associated with genetic clusters. The residual support relative to row hypotheses. The values in the table are the for STRUCTURE groups may reflect patterns of past difference in partial Mantel r for test 1 (genetic distance * column hypothesis | row hypothesis) and test 2 (genetic population fission/fusion and expansion from refugia distance * row hypothesis | column hypothesis). Positive over long time scales (e.g., Dyer et al. 2010; Cushman values indicate support for the column hypothesis relative to 2015). For example, Li et al. (2012) studied effects of the row hypothesis. The matrix shows full support for past climatic events on intraspecific divergence and hypothesis g15, precipitation seasonality, and no other hypotheses are supported range shifts in Reamuria soongorica using chloroplast DNA from 27 natural populations across western China. They found eight haplotypes clustered into accounted for by this landscape resistance model that three clades divided between eastern and western seems to be associated with three or seven genetic china. They surmised that the resulting pattern indi- clusters. cated past regional range expansion corresponding to the uplift of the Qinghai-Tibet Plateau and development of desert ecosystems during the last glacial age. Discussion Our analysis is unable to test that hypothesis, but our observation of residual structure most strongly asso- Isolation by Climatic Gradients ciated with three genetic clusters could plausibly reflect the pattern of population differentiation de- Our results indicated strong genetic differentiation of scribed by Li et al. (2012). Like Cushman et al. (2014), R.soongorica across the Mongolian Plateau, and that we found both strong clinal patterns in genetic this genetic differentiation was primarily clinal in differentiation related to climatic gradients and also nature, but also had some attributes independently independent underlying discrete patterns of genetic associated with Bayesian clustering. Gene flow among clustering simultaneously exist across the population. populations is quite high (Nm = 2.0899), but consid- Qian et al. (2008) used ISSR markers across seven erable genetic differentiation still exists among subpopulations of Reamuria soongorica in a relatively populations (FST = 0.26233), consistent with gene small area in western China. Similar to our results, flow along a cline. STRUCTURE analysis revealed Qian et al. (2008) found relatively strong genetic weak clustering with high degree of admixture, also structure coupled with relatively high gene flow. They suggesting a clinal gradient of gene flow across the analyzed isolation by distance with Mantel tests and study area. Schwartz and McKelvey (2009) found that found strong correlation between genetic differen- Bayesian clustering approaches, such as STRUC- tiation and geographical distance. Our results provide TURE, detect spurious clusters when sampling is a next step beyond those of Qian et al. (2008) in that spatially segregated along a genetic cline. Our sam- we directly competed 21 alternative hypotheses, pling was spatially distributed across several hundred including isolation by distance and STRUCTURE 123 934 Plant Ecol (2015) 216:925–937 groupings, and found that isolation by distance was not The relationship between seasonal variation in independently supported and that the dominant drivers precipitation and gene flow observed in this study and of gene flow are related to climatic gradients. in the Yang et al. (2013) study may be driven by the Interestingly, we found that variation in precipita- effects of climate on phenology. Climate conditions, tion is a much stronger predictor of genetic differen- especially precipitation and temperature, will affect tiation among populations of R. soongorica than is growth and phenology. For flowering plants in the variation in temperature. This suggests that gradients temperate zone, if it is warm with ample rain in the of changing precipitation may act as resistant factors spring, plants will leave dormancy and enter active limiting gene flow, and provide stronger prediction of growth earlier than in cold and dry locations. If genetic differentiation than isolation by distance, and precipitation is low in summer and autumn, the period that variation in temperature has little relation to gene of active growth will terminate early and plants will flow. A very similar result was found for Populus enter dormancy earlier than in areas with higher fremontii by Cushman et al. (2014). They found no summer and autumn precipitation. Likewise, pre- support for genetic differentiation based on differ- cipitation in autumn and winter can significantly affect ences between climatic zones or point climatic the timing of the next year’s bud-burst and flowering. conditions at the sites of populations, but strong Specifically, if precipitation of the preceding autumn support for increased genetic differentiation as a and winter was high, bud-burst and flowering will both function of cumulative difference in winter and spring occur earlier in the year than average (Bai et al. 2010; precipitation between populations. This suggests that Guo et al. 2012; Song et al. 2012a, b). seasonal differences in precipitation result in reduced Across the Mongolian Plateau, there is very little gene flow, plausibly due to the effects on flowering precipitation in spring. Therefore, differences in phenology. We posit that the same process may be winter and summer precipitation likely have very occurring in R. soongorica across the Mongolian strong effects on the ecophysiological factors control- Plateau, with differences in seasonal precipitation ling flowering and other reproductive processes. Our driving differences in flowering phenology, leading to results are consistent with the hypothesis that differ- cumulative isolation by climatic gradients. ences in flowering phenology along gradients of Yang et al. (2013) found similar relationships changing precipitation seasonality may influence between gene flow and variation in seasonal precipita- pollination and seed diffusion, and drive differential tion for a clade of Caragana species on the Ordos patterns of gene flow. In addition, the differential Plateau, which is a subregion of the current study. That climatic conditions across the precipitation gradient study found that gene flow within and among Cara- may impose marked local directional selection on the gana populations was driven primarily by winter local populations (Yang et al. 2013). These climatic precipitation and precipitation seasonality, with no differences likely reduce fitness of locally maladapted independent relationship with other climate gradients individuals, resulting in population divergence and or isolation by distance. The results in this study are maintenance of reproductive isolation (de Leoń et al. similar, except that we found no support for a 2010; Gavrilets 2000; Gavrilets and Vose 2007; relationship between gradients of winter precipitation Niemiller et al. 2008; Nosil 2008). and gene flow, and only found that variation in seasonal precipitation across space acted as a resistant Patterns of genetic diversity factor attenuating gene flow. The general similarity of the results suggests that seasonal variation of pre- Our genetic analysis revealed that R. soongorica cipitation may act as a structuring factor limiting gene populations across the Inner Mongolian plateau flow for a range of plant taxa across the Mongolian exhibit abundant genetic diversity. The amount of Plateau, and that other climate factors seem to play a genetic diversity in plant populations reflects the result lesser role. This knowledge could be of great value in of evolutionary actions, including accumulation of projecting potential effects of climate change on adaptive genetic changes to environmental gradients. population migration and gene flow of R. soongorica High genetic diversity is associated with greater and potentially other plant species as well. evolutionary potential and the ability for breeding

123 Plant Ecol (2015) 216:925–937 935 and genetic improvement (Beebee and Rowe 2003; resistance), but this model leaves a portion of variance Chai et al. 2010; Jiang 2004). Such high genetic in spatial genetic structure unexplained. This residual diversity is not unexpected given that R. soongorica is variation is related to other factors not evaluated in the widely distributed across typical desert, steppe desert, model, but is correlated with the STRUCTURE desert steppe, and typical steppe in the region. groups, and may reflect patterns of past population For all measures of genetic diversity assessed, we fission/fusion and expansion from refugia over long found the highest genetic diversity of the species was in time scales (e.g., Dyer et al. 2010), as described for the western desert habitats. This was expected, as it is this species in Li et al. (2012). generally to be expected that genetic diversity will be We demonstrated that genetic differentiation of R. highest in the main distribution area of a species’ range soongorica across the Mongolian Plateau is clinal in where effective population size is highest, and de- nature and is driven primarily by differential land- crease toward the periphery or in areas where the range scape resistance across areas with changing patterns of is expanding or populations are fluctuating widely. The seasonal precipitation. Changing seasonality of pre- lower genetic diversity in the middle and eastern parts cipitation probably limits gene flow by affecting of the range suggest perhaps that populations of R. phenology of flowering events that lead to increased soongorica may have fluctuated or that local migration prezygotic isolation among populations. Finding that may have occurred that has reduced genetic diversity in seasonal patterns of precipitation, and not tem- these areas, and that conversely, the populations in the perature, drive population connectivity, and gene flow far western part of the study area may have had a longer may have important implications for predicting the stable period with relatively high effective population effects of climate change on this keystone foundation size. Understanding the causes of this expected pattern species and devising effective strategies to utilize it in of genetic diversity could be important to piecing restoration efforts to ameliorate ongoing desertifica- together the ecological history of the species in the tion in the region. For example, further research would Mongolian Plateau and anticipating the effects of be warranted to determine if the climatic differen- climate change, and effectively using the species to tiation we observe is associated with any adaptive mitigate desertification. differences among populations. Such adaptive differences would be essential to guide selection of seed sources for restoration plantings that are adapted to the Conclusion climate of the target environment. In addition, understanding the climatic controls on gene flow across the Our modeling results do not support isolation by Mongolian Plateau could help guide conservation and distance as a significant factor in determining genetic land management by clarifying how naturally con- differentiation in R. soongorica after accounting for nected populations are and where gene flow is most landscape resistance due to climate gradients. attenuated. This, in turn, could guide restoration STRUCTURE clustering identifies grouping of ge- efforts to maintain gene flow by placing stepping netically similar populations without any a priori stone populations across areas of steep climatic hypotheses of driving factors. Given they lack any a change to enable gene flow through these highly priori basis these clusters are observations of differ- resistant zones. entiation and not explanations. We treated isolation by We found that genetic diversity was highest in the STRUCTURE clustering groupings as null models in western part of the sampled population, perhaps this analysis, and our finding that there is residual indicating that this region has historically harbored independent support for them after accounting for the highest effective population size of the species or isolation by distance and isolation by resistances may have served as the source of recent range suggests that these patterns are not merely spurious expansion to other parts of the sampled range which results of clustered sampling along clinal population exhibited lower genetic diversity. Understanding the gradients (Schwartz and McKelvey 2009; Cushman ecological drivers of these relationships might be and Landguth 2010). Genetic structure of R. soon- critical to resolving the causes of the geographical gorica population is primarily related to gradients of pattern of diversity, and could be important in increasing isolation by climatic differences (landscape understanding the ecology of the species sufficiently 123 936 Plant Ecol (2015) 216:925–937 to anticipate climate change effects and effectively Dyer RJ, Nason JD, Garrick RC (2010) Landscape modeling of implement management strategies to restore the gene flow: improved power using conditional genetic distance derived from topology of population networks. Mol species and combat desertification. Ecol 19:3746–3759 Editorial Committee of Flora Reipublicae Popular Sinicae, Acknowledgments This work was supported by The State Chinese Academy of Science (1990) Flora of China (50 vol Key Basic Research Development Programme of China (Grant part 2). Science Press, Beijing No. 2012CB722201),science and technology research projects Evanno G, Regnaut S, Goudet J (2005) Detecting the number of of Inner Mongolia Education Department of China (NJ10002), clusters of individuals using the software structure: a The Natural Science Foundation of Inner Mongolia simulation study. Mol Ecol 14:2611–2620 Autonomous Region of China (2014MS0350), and US Forest Excoffer L, Smouse PE, Quatrio JM (1992) Analysis of mole- Service Rocky Mountain Research Station. cular variance inferred from metric distances among DNA haplotypes: application to human mitochondrial DNA re- striction data [J]. Genetics 131:479–2491 Falush D, Stephens M, Pritchard JK (2003) Inference of population structure using multilocus genotype data: References linked loci and correlated allele frequencies. Genetics 164:1567–1587 Amos J, Bennet AF, Mac-Nally R, Newell G, Radford JQ, Gavrilets S (2000) Waiting time to parapatric speciation. Proc R Pavlova A, Thompson J, White M, Sunnucks P (2012) Soc B 570(267):2483–2492 Predicting landscape genetic consequences of habitat loss, Gavrilets S, Vose A (2007) Case studies and mathematical fragmentation and mobility for species of woodland birds. models of ecological speciation. 2. Palms on an oceanic PLoS One 7:e30888 island. Mol Ecol 16:2910–2921 Bai J, Ge QS, Dai JH, Wang Y (2010) Relationship between Guo ZX, Zhang XN, Wang ZM et al (2012) Responses of woody plants phenology and climate factors in Xi’an, vegetation phenology in Northeast China to climate China. Chin J Plant Ecol 34(11):1274–1282 change. Chin J Ecol 29(3):578–585 Beebee T, Rowe G (2003) An introduction to molecular ecol- Gupta M, Chyi YS, Romero-Severson J, Owen JL (1994) Am- ogy. Oxford University Press, New York plification of DNA markers from evolutionarily diverse Chai XY, Chen SL, Xu W (2010) Using inter-simple sequence genomes using single primers of simple-sequence repeats. repeat markers to analyze the genetic structure of natural Theor Appl Genet 89:998–1006 Pteroceltis tatarinowii populations and implications for Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A (2005) species conservation. Plant Syst Evol 285:65–73 Very high-resolution interpolated climate surfaces for Coyne JA, Orr AH (2004) Speciation. Sinauer Associates, global land areas. Int J Climatol 25:1965–1978 Sunderland Hou XY (1983) Vegetation of China with reference to its geo- Cui DF (1988) A primary study on the classification, distribu- graphical distribution. Ann Mo Bot Gard 70:509–549 tion, ecological characteristics, morphological character- Inner Mongolia and Ningxia Comprehensive expedition of istics and anatomical characteristics of genus Reaumuria in Chinese academy of sciences (1985) Vegetation of Inner China. Arid Zone Res 1:65–69 Mongolia. Science Press, Beijing Cushman SA (2015) Pushing the envelope in genetic analysis of Iverson LR, Prasad AM (2002) Potential redistribution of tree species invasion. Mol Ecol 24:259–262 species habitat under five climate change scenarios in the Cushman SA, Landguth EL (2010) Spurious correlations and eastern US. For Ecol Manag 155:205–222 inference in landscape genetics. Mol Ecol 19:3592–3602 Jiang HQ (2004) Plants ecology. Higher Education Press, Cushman SA, McKelvey KS, Hayden J, Schwartz MK (2006) Beijing Gene-flow in complex landscapes: testing multiple models Legendre P, Fortin MJ (2010) Comparison of the Mantel test and with causal modeling. Am Nat 168:486–499 alternative approaches for detecting complex multivariate Cushman SA, Wasserman TN, Landguth E, Shirk AJ (2013) Re- relationships in the spatial analysis of genetic data. Mol evaluating causal modeling with Mantel tests in landscape Ecol 10:831–844 genetics. Diverstiy 5:51–72. doi:10.3390/d50x000x Li B (1990) Natural resources and environment research in Cushman SA, Max T, Meneses N, Evans LM, Ferrier S, Hon- Ordos Plateau, Inner Mongolia. Science Press, Beijing chak B, Whitham TG, Allan GJ (2014) Landscape genetic Li JP, Yang XG, Fu H, Zhang BL (2005a) The content and connectivity in a riparian foundation tree is jointly driven distribution characteristics of some osmotic adjusting ma- by climatic gradients and river networks. Ecol Appl terials in three species of desert plants in Alashan desert of 24:1000–1014. doi:10.1890/13-1612.1 northwest China. Pratacult Sci 22(9):35–38 de Leoń LF, Bermingham E, Podos J, Hendry AP (2010) Li X, Wang YC, Zheng R (2005b) Water parameters of xeric Divergence with gene flow as facilitated by ecological shrubs in west Erdos region (I). J Desert Res 25(4):581–586 differences: within-island variation in Darwin’s finches. Li XL, Chen J, Wang G (2008) Spatial autocorrelation analysis of Philos Trans R Soc B 365:1041–1052 ISSR genetic variation of Reaumuria soongorica population Doyle J (1999) DNA protocols for plants—CTAB total DNA in northwest of China. J Desert Res 28(3):469–472 isolation. In: Hewitt GM, Johnston A (eds) Molecular Li Z, Chen J, Zhao G-F, Guo Y-P, Kou Y-X, Ma Y-Z, Wang G, techniques in taxonomy. Springer, Berlin, pp 283–293 Ma X-F (2012) Response of a desert shrub to past

123 Plant Ecol (2015) 216:925–937 937

geological and climatic change: a phylogeographic study Song WM, Zhou HY, Jia RL, Zhao X, Feng L, Tan HJ (2008) of Reamuria soongorica (Tabmaricaceae) in western Chi- Response of photosynthesis function and trehalose content na. J Syst Evol 50:351–361 of four desert plants to gradual drought stress. J Desert Res Liu JQ, Qiu MX, Pu JC, Lu ZM (1982) The typical extreme 28:449–454 (in Chinese) xerophyte —Reaumuria soongorica in the desert of China. Song CQ, You SC, Ke LH, Liu GH, Zhong XK (2012a) Acta Bot Sin 24(5):485–488 Phenological variation of typical vegetation types in Ma YQ (1989) Flora of Inner Mongolia, vol 3, 2nd edn. The northern Tibet and its response to climate changes. Acta Peoples Press of Inner Mongolia, Hohhot Ecol Sin 32(4):1045–1055 Ma MH, Kong LS (1998) The bio-ecological characteristics of Song XM, Yang JY, Lv MT, Yang M, Zhang ZR (2012b) Re- Reaumuria soongorica on the border of oasis at Hutubi, sponses of Reaumuria soongorica seed germination to salt Xinjiang. Acta Phytoecol Sin 22(3):237–244 stress and moderate drought. J Desert Res 32(6):1674–1680 Mantel N (1967) The detection of disease clustering and a Wang YR, Zeng YJ, Fu H, Chen SK (2002a) Affects of over generalized regression approach. Cancer Res 27:209–220 grazing and enclosure on desert vegetation succession of McKinnon JS, Mori S, Blackman BK, David L, Kingsley DM Reaumuria soongrica. J Desert Res 22(4):321–327 et al (2004) Evidence for ecology’s role in speciation. Wang YR, Zeng YJ, Zhang BL, Ta LT (2002b) Water distri- Nature 429:294–298 bution patterns in different degraded desert grasslands of Meirmans PG (2012) The trouble with isolation by distance. Reaumurta soongorica. Chin J Appl Ecol 13(8):962–966 Mol Ecol 21:2839–2846 Wasserman TN, Cushman SA, Schwartz MK, Wallin DO (2010) Nei M (1972) Genetic distance between populations. Am Nat Spatial scaling and multi-model inference in landscape 106:282–292 genetics: Martes americana in northern Idaho. Landsc Ecol Nei M (1973) Analysis of gene diversity in subdivided 25:1601–1612 populations. Proc Natl Acad Sci USA 70:3321–3323 Weising K, Atkinson RG, Gardner RC (1995) Genomic fin- Niemiller ML, Fitzpatrick BM, Miller BT (2008) Recent di- gerprinting by microsatellite-primed PCR: a critical eval- vergence-with-gene-flow in Tenessee cave salamanders uation. PCR Methods Appl 4:249–255 (Plethodontidae; Gyrinophylus) inferred from gene ge- Wright S (1931) Statistical methods in biology. J Am Stat Assoc nealogies. Mol Ecol 17:2258–2275 26(173):155–163 Nix H (1986) A biogeographic analysis of Australian elapid Wu ZY (1980) Vegetation of China. Science Press, Beijing snakes. In: Atlas of elapid snakes of Australia. Australian Xiao SC, Xiao HL, Song YX, Duan ZH, Lu MF (2006) Den- Government Publishing Service, Canberra drochronology study on response of Reaumurta soongorica Nosil P (2008) Speciation with gene flow could be common. to water-heat variation. J Desert Res 26(4):548–552 Mol Ecol 17:2103–2106 Yang JY, Cushman SA, Yang J, Yang MB, Bao TJ (2013) Ef- Pritchard JK, Stephens P, Donnelly P (2000) Inference of fects of climatic gradients on genetic diversification of population structure using multilocus genotype data. Ge- Caragana on the Ordos plateau, China. Landsc Ecol 28(9): netics 155:945–959 1729–1741 Qian W, Ge S, Hong DY (2001) Genetic variation within and Yeh FC, Yang RC, Boyle T (1999) POPGENE. Microsoft among populations of a wild rice Oryza granulate from window-based freeware for population genetic analysis. China detected by RAPD and ISSR markers. Theor Appl Release 1.31. University of Alberta, Edmonton Genet 102:440–449 Yong SP (1990) Study on the basic characteristics of desert Qian Z-Q, Xu L, Wang Y-L, Yang J, Zhao G-F (2008) Eco- flora. J Inn Mongolia Univ 21(2):241–247 logical genetics of Reamuria soongorica (Pall.) Maxim. Zeng YJ, Wang YR, Zhuang GH, Yang ZS (2004) Seed ger- population in the oasis-desert ecotone in Fukang, Xinjiang, mination responses of Reaumuria soongorica and Zygo- and its implications for molecular evolution. Biochem Syst phyllum xanthoxylum to drought stress and sowing depth. Ecol 36:593–601 Acta Ecol Sin 24(8):1630–1634 Rehfeldt GE, Crookston NL, Warwell MV, Evans JS (2006) Zhang DY (2005) Discuss on some systematical problems of Empirical analyses of plant-climate relationships for the Tamaricaceae. Acta Bot Yunnanica 27(5):471–478 western United States. Int J Plant Sci 167(6):1123–1150 Zhang YJ, Gao RX, Li QF (2008) Analysis on the population Schwartz MK, McKelvey KS (2009) Why sampling scheme genetic diversity of desert shrub Reaumuria soongorica in matters: the effect of sampling scheme on landscape disturbed habitats. J Arid Land Resour Environ 22(3): genetic results. Conserv Genet 10(2):441–452 147–151 Smouse PE, Long JC, Sokal RR (1986) Multiple regression and Zietkiewicz E, Rafalski A, Labuda D (1994) Genome finger- correlation extensions of the Mantel test of matrix corre- printing by simple sequence repeat (SSR)-anchored poly- spondence. Syst Zool 35:627–632 merase chain reaction amplification. Genomics 20:176–183

123